src/bin/lttng-relayd/health-relayd.c

   1 /*
   2  * Copyright (C) 2013 - Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
   3  *
   4  * This program is free software; you can redistribute it and/or modify
   5  * it under the terms of the GNU General Public License, version 2 only,
   6  * as published by the Free Software Foundation.
   7  *
   8  * This program is distributed in the hope that it will be useful,
   9  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  11  * GNU General Public License for more details.
  12  *
  13  * You should have received a copy of the GNU General Public License along
  14  * with this program; if not, write to the Free Software Foundation, Inc.,
  15  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
  16  */
  17
  18 #define _LGPL_SOURCE
  19 #include <fcntl.h>
  20 #include <getopt.h>
  21 #include <grp.h>
  22 #include <limits.h>
  23 #include <pthread.h>
  24 #include <signal.h>
  25 #include <stdio.h>
  26 #include <stdlib.h>
  27 #include <string.h>
  28 #include <sys/ipc.h>
  29 #include <sys/resource.h>
  30 #include <sys/shm.h>
  31 #include <sys/socket.h>
  32 #include <sys/stat.h>
  33 #include <sys/types.h>
  34 #include <urcu/list.h>
  35 #include <poll.h>
  36 #include <unistd.h>
  37 #include <sys/mman.h>
  38 #include <assert.h>
  39 #include <urcu/compiler.h>
  40 #include <inttypes.h>
  41
  42 #include <common/defaults.h>
  43 #include <common/common.h>
  44 #include <common/consumer/consumer.h>
  45 #include <common/consumer/consumer-timer.h>
  46 #include <common/compat/poll.h>
  47 #include <common/sessiond-comm/sessiond-comm.h>
  48 #include <common/utils.h>
  49 #include <common/compat/getenv.h>
  50 #include <common/fd-tracker/utils.h>
  51
  52 #include "lttng-relayd.h"
  53 #include "health-relayd.h"
  54
  55 /* Global health check unix path */
  56 static
  57 char health_unix_sock_path[PATH_MAX];
  58
  59 int health_quit_pipe[2];
  60
  61 /*
  62  * Check if the thread quit pipe was triggered.
  63  *
  64  * Return 1 if it was triggered else 0;
  65  */
  66 static
  67 int check_health_quit_pipe(int fd, uint32_t events)
  68 {
  69         if (fd == health_quit_pipe[0] && (events & LPOLLIN)) {
  70                 return 1;
  71         }
  72
  73         return 0;
  74 }
  75
  76 /*
  77  * Send data on a unix socket using the liblttsessiondcomm API.
  78  *
  79  * Return lttcomm error code.
  80  */
  81 static int send_unix_sock(int sock, void *buf, size_t len)
  82 {
  83         /* Check valid length */
  84         if (len == 0) {
  85                 return -1;
  86         }
  87
  88         return lttcomm_send_unix_sock(sock, buf, len);
  89 }
  90
  91 static int create_lttng_rundir_with_perm(const char *rundir)
  92 {
  93         int ret;
  94
  95         DBG3("Creating LTTng run directory: %s", rundir);
  96
  97         ret = mkdir(rundir, S_IRWXU);
  98         if (ret < 0) {
  99                 if (errno != EEXIST) {
 100                         ERR("Unable to create %s", rundir);
 101                         goto error;
 102                 } else {
 103                         ret = 0;
 104                 }
 105         } else if (ret == 0) {
 106                 int is_root = !getuid();
 107
 108                 if (is_root) {
 109                         ret = chown(rundir, 0,
 110                                         utils_get_group_id(tracing_group_name));
 111                         if (ret < 0) {
 112                                 ERR("Unable to set group on %s", rundir);
 113                                 PERROR("chown");
 114                                 ret = -1;
 115                                 goto error;
 116                         }
 117
 118                         ret = chmod(rundir,
 119                                         S_IRUSR | S_IWUSR | S_IXUSR | S_IRGRP | S_IXGRP | S_IROTH | S_IXOTH);
 120                         if (ret < 0) {
 121                                 ERR("Unable to set permissions on %s", health_unix_sock_path);
 122                                 PERROR("chmod");
 123                                 ret = -1;
 124                                 goto error;
 125                         }
 126                 }
 127         }
 128
 129 error:
 130         return ret;
 131 }
 132
 133 static
 134 int parse_health_env(void)
 135 {
 136         const char *health_path;
 137
 138         health_path = lttng_secure_getenv(LTTNG_RELAYD_HEALTH_ENV);
 139         if (health_path) {
 140                 strncpy(health_unix_sock_path, health_path,
 141                         PATH_MAX);
 142                 health_unix_sock_path[PATH_MAX - 1] = '\0';
 143         }
 144
 145         return 0;
 146 }
 147
 148 static
 149 int setup_health_path(void)
 150 {
 151         int is_root, ret = 0;
 152         char *home_path = NULL, *rundir = NULL, *relayd_path = NULL;
 153
 154         ret = parse_health_env();
 155         if (ret) {
 156                 return ret;
 157         }
 158
 159         is_root = !getuid();
 160
 161         if (is_root) {
 162                 rundir = strdup(DEFAULT_LTTNG_RUNDIR);
 163                 if (!rundir) {
 164                         ret = -ENOMEM;
 165                         goto end;
 166                 }
 167         } else {
 168                 /*
 169                  * Create rundir from home path. This will create something like
 170                  * $HOME/.lttng
 171                  */
 172                 home_path = utils_get_home_dir();
 173
 174                 if (home_path == NULL) {
 175                         /* TODO: Add --socket PATH option */
 176                         ERR("Can't get HOME directory for sockets creation.");
 177                         ret = -EPERM;
 178                         goto end;
 179                 }
 180
 181                 ret = asprintf(&rundir, DEFAULT_LTTNG_HOME_RUNDIR, home_path);
 182                 if (ret < 0) {
 183                         ret = -ENOMEM;
 184                         goto end;
 185                 }
 186         }
 187
 188         ret = asprintf(&relayd_path, DEFAULT_RELAYD_PATH, rundir);
 189         if (ret < 0) {
 190                 ret = -ENOMEM;
 191                 goto end;
 192         }
 193
 194         ret = create_lttng_rundir_with_perm(rundir);
 195         if (ret < 0) {
 196                 goto end;
 197         }
 198
 199         ret = create_lttng_rundir_with_perm(relayd_path);
 200         if (ret < 0) {
 201                 goto end;
 202         }
 203
 204         if (is_root) {
 205                 if (strlen(health_unix_sock_path) != 0) {
 206                         goto end;
 207                 }
 208                 snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
 209                         DEFAULT_GLOBAL_RELAY_HEALTH_UNIX_SOCK,
 210                         (int) getpid());
 211         } else {
 212                 /* Set health check Unix path */
 213                 if (strlen(health_unix_sock_path) != 0) {
 214                         goto end;
 215                 }
 216
 217                 snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
 218                         DEFAULT_HOME_RELAY_HEALTH_UNIX_SOCK,
 219                         home_path, (int) getpid());
 220         }
 221
 222 end:
 223         free(rundir);
 224         free(relayd_path);
 225         return ret;
 226 }
 227
 228 static
 229 int open_unix_socket(void *data, int *out_fd)
 230 {
 231         int ret;
 232         const char *path = data;
 233
 234         ret = lttcomm_create_unix_sock(path);
 235         if (ret < 0) {
 236                 goto end;
 237         }
 238
 239         *out_fd = ret;
 240         ret = 0;
 241 end:
 242         return ret;
 243 }
 244
 245 /*
 246  * Thread managing health check socket.
 247  */
 248 void *thread_manage_health(void *data)
 249 {
 250         int sock = -1, new_sock = -1, ret, i, pollfd, err = -1;
 251         uint32_t revents, nb_fd;
 252         struct lttng_poll_event events;
 253         struct health_comm_msg msg;
 254         struct health_comm_reply reply;
 255         int is_root;
 256         char *sock_name;
 257
 258         DBG("[thread] Manage health check started");
 259
 260         setup_health_path();
 261
 262         rcu_register_thread();
 263
 264         /* We might hit an error path before this is created. */
 265         lttng_poll_init(&events);
 266
 267         /* Create unix socket */
 268         ret = asprintf(&sock_name, "Unix socket @ %s", health_unix_sock_path);
 269         if (ret == -1) {
 270                 PERROR("Failed to allocate unix socket name");
 271                 err = -1;
 272                 goto error;
 273         }
 274         ret = fd_tracker_open_unsuspendable_fd(the_fd_tracker, &sock,
 275                         (const char **) &sock_name, 1, open_unix_socket,
 276                         health_unix_sock_path);
 277         free(sock_name);
 278         if (ret < 0) {
 279                 ERR("Unable to create health check Unix socket");
 280                 err = -1;
 281                 goto error;
 282         }
 283
 284         is_root = !getuid();
 285         if (is_root) {
 286                 /* lttng health client socket path permissions */
 287                 ret = chown(health_unix_sock_path, 0,
 288                                 utils_get_group_id(tracing_group_name));
 289                 if (ret < 0) {
 290                         ERR("Unable to set group on %s", health_unix_sock_path);
 291                         PERROR("chown");
 292                         err = -1;
 293                         goto error;
 294                 }
 295
 296                 ret = chmod(health_unix_sock_path,
 297                                 S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP);
 298                 if (ret < 0) {
 299                         ERR("Unable to set permissions on %s", health_unix_sock_path);
 300                         PERROR("chmod");
 301                         err = -1;
 302                         goto error;
 303                 }
 304         }
 305
 306         /*
 307          * Set the CLOEXEC flag. Return code is useless because either way, the
 308          * show must go on.
 309          */
 310         (void) utils_set_fd_cloexec(sock);
 311
 312         ret = lttcomm_listen_unix_sock(sock);
 313         if (ret < 0) {
 314                 goto error;
 315         }
 316
 317         /* Size is set to 1 for the consumer_channel pipe */
 318         ret = lttng_poll_create(&events, 2, LTTNG_CLOEXEC);
 319         if (ret < 0) {
 320                 ERR("Poll set creation failed");
 321                 goto error;
 322         }
 323
 324         ret = lttng_poll_add(&events, health_quit_pipe[0], LPOLLIN);
 325         if (ret < 0) {
 326                 goto error;
 327         }
 328
 329         /* Add the application registration socket */
 330         ret = lttng_poll_add(&events, sock, LPOLLIN | LPOLLPRI);
 331         if (ret < 0) {
 332                 goto error;
 333         }
 334
 335         lttng_relay_notify_ready();
 336
 337         while (1) {
 338                 DBG("Health check ready");
 339
 340                 /* Inifinite blocking call, waiting for transmission */
 341 restart:
 342                 ret = lttng_poll_wait(&events, -1);
 343                 if (ret < 0) {
 344                         /*
 345                          * Restart interrupted system call.
 346                          */
 347                         if (errno == EINTR) {
 348                                 goto restart;
 349                         }
 350                         goto error;
 351                 }
 352
 353                 nb_fd = ret;
 354
 355                 for (i = 0; i < nb_fd; i++) {
 356                         /* Fetch once the poll data */
 357                         revents = LTTNG_POLL_GETEV(&events, i);
 358                         pollfd = LTTNG_POLL_GETFD(&events, i);
 359
 360                         if (!revents) {
 361                                 /* No activity for this FD (poll implementation). */
 362                                 continue;
 363                         }
 364
 365                         /* Thread quit pipe has been closed. Killing thread. */
 366                         ret = check_health_quit_pipe(pollfd, revents);
 367                         if (ret) {
 368                                 err = 0;
 369                                 goto exit;
 370                         }
 371
 372                         /* Event on the registration socket */
 373                         if (pollfd == sock) {
 374                                 if (revents & LPOLLIN) {
 375                                         continue;
 376                                 } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) {
 377                                         ERR("Health socket poll error");
 378                                         goto error;
 379                                 } else {
 380                                         ERR("Unexpected poll events %u for sock %d", revents, pollfd);
 381                                         goto error;
 382                                 }
 383                         }
 384                 }
 385
 386                 new_sock = lttcomm_accept_unix_sock(sock);
 387                 if (new_sock < 0) {
 388                         goto error;
 389                 }
 390
 391                 /*
 392                  * Set the CLOEXEC flag. Return code is useless because either way, the
 393                  * show must go on.
 394                  */
 395                 (void) utils_set_fd_cloexec(new_sock);
 396
 397                 DBG("Receiving data from client for health...");
 398                 ret = lttcomm_recv_unix_sock(new_sock, (void *)&msg, sizeof(msg));
 399                 if (ret <= 0) {
 400                         DBG("Nothing recv() from client... continuing");
 401                         ret = close(new_sock);
 402                         if (ret) {
 403                                 PERROR("close");
 404                         }
 405                         new_sock = -1;
 406                         continue;
 407                 }
 408
 409                 rcu_thread_online();
 410
 411                 assert(msg.cmd == HEALTH_CMD_CHECK);
 412
 413                 memset(&reply, 0, sizeof(reply));
 414                 for (i = 0; i < NR_HEALTH_RELAYD_TYPES; i++) {
 415                         /*
 416                          * health_check_state return 0 if thread is in
 417                          * error.
 418                          */
 419                         if (!health_check_state(health_relayd, i)) {
 420                                 reply.ret_code |= 1ULL << i;
 421                         }
 422                 }
 423
 424                 DBG2("Health check return value %" PRIx64, reply.ret_code);
 425
 426                 ret = send_unix_sock(new_sock, (void *) &reply, sizeof(reply));
 427                 if (ret < 0) {
 428                         ERR("Failed to send health data back to client");
 429                 }
 430
 431                 /* End of transmission */
 432                 ret = close(new_sock);
 433                 if (ret) {
 434                         PERROR("close");
 435                 }
 436                 new_sock = -1;
 437         }
 438
 439 error:
 440         lttng_relay_stop_threads();
 441 exit:
 442         if (err) {
 443                 ERR("Health error occurred in %s", __func__);
 444         }
 445         DBG("Health check thread dying");
 446         unlink(health_unix_sock_path);
 447         if (sock >= 0) {
 448                 ret = fd_tracker_close_unsuspendable_fd(the_fd_tracker, &sock,
 449                                 1, fd_tracker_util_close_fd, NULL);
 450                 if (ret) {
 451                         PERROR("close");
 452                 }
 453         }
 454
 455         /*
 456          * We do NOT rmdir rundir nor the relayd path because there are
 457          * other processes using them.
 458          */
 459
 460         lttng_poll_clean(&events);
 461
 462         rcu_unregister_thread();
 463         return NULL;
 464 }